LUA Progress bar

[Adriweb]

I know that the code is not optimized (not checking double will be a nice fix for sure).

See his edit above.
A note about that : in general, memoization (yes that's the right word) is a good idea.
(Edit : well, I mean, rather about "caching" results instead of re-calculating stuff several times - if you can't do it otherwise, of course)

Though many o the constants are declared (as local) outside the code snippet i shared (program is currently 700 lines + 8000 lines database)

Using locals instead of globals whenever you can is better, in short.
Also valid with functions from modules like math etc. : when you use them (especially a lot, like in loops, localize them before, it'll be faster in the loop. See above code with sqrt)

I'm not familiar with the coroutine 'function', but I hope that it will address the problem.

That sure is advanced Lua and will likely be confusing for you if you're only beginning - but don't worry, it's not that scary

For more optimization tricks : archives_voir.php?id=6720
(and there is another one more recent but I can't access it because the host (free.fr) recently changed its policy on listing the content of ftp directory... meh)

[Jens_K]

Glad I could help!
Coroutines are basically threads in lua, altough I don't think real multithreading is possible on the nspire (?). But they are useful when you want to pause and resume functions

(and there is another one more recent but I can't access it because the host (free.fr) recently changed its policy on listing the content of ftp directory... meh)

I found the newer (better) version in my download-folder, so I attached it (I hope that's ok because of copyrights'n'stuff)

[Adriweb]

Yep, thanks
It's actually "better" for more advanced people - but the one I linked would probably be more understandable to beginners though (which was who it was made for originally, too)

I'll make an index.html with links to the other stuff lying in the folder when I have an FTP access to the server.

[FrederikTheisen]

I was about to post this but it might not be a problem anymore. (yay)

I have got a problem with the way it works, which I'm not able to solve.
The program continues past the coroutine before running the code in the coroutine.

It's probably because I don't understand how it works.

As you can see here I need the maxDist immediatly after I calculate it

Code: Tout sélectionner

         maxDist = 0
         --getMaxDist()
         CRgetMaxDist = coroutine.create(getMaxDist)
         -- SET MOLECULE SIZE (DISTANCE)
         cDist = 350+maxDist^1.55
      end
   end
   --dataLoad = false
end


I solved the problem by finishing the remaining tasks inside the 'on.timer' event when the progress = 1

Code: Tout sélectionner

if dataLoadPRG == 1 then
   timer.stop()
   dataLoad = false
   data = true
   cDist = 350+maxDist^1.55
   getRotation()
end

It seems to work ok, but it still takes an awfully long time for large molecules (that is what I'm working with ).

[Adriweb]

Try without the coroutine (direct call) but do the window invalidate every {x} loop index ? for example, something like if index%100 == 0 then platform.window:invalidate() end

[Jens_K]

Or do something like this:

Code: Tout sélectionner

function getMaxDist()
   local sqrt, getMilliSecCounter = math.sqrt, timer.getMilliSecCounter
   local dist, now
   local nAtoms = molTable["nAtoms"]
   for i = 1,nAtoms-1 do
      for j = i+1,nAtoms do
         dist = sqrt((molTable[i]["x"] - molTable[j]["x"])^2  +  (molTable[i]["y"] - molTable[j]["y"])^2  +  (molTable[i]["z"] - molTable[j]["z"])^2)
         if dist > maxDist then
            maxDist = dist
         end
      end
      now=getMilliSecCounter()
      if now>=lastUpdate+updateInterval then
         coroutine.yield(i/(nAtoms-1))
         lastUpdate=now
      end
   end
   coroutine.yield(1)
end

local lastUpdate=timer.getMilliSecCounter()
local updateInterval=1000   --ms
CRgetMaxDist=coroutine.create(getMaxDist)
local progress=0
local finished=false

I don't know how fast the OS gets the system clock's value though, so it might be a bit slower.

[FrederikTheisen]

Quick test with Jens_Ks proposed change.

The calculator (measured with stopwatch) is about 2 times faster. 50s vs 26s

The PC software (measured using timer.get...) is about 70 times faster. 8444ms vs 122ms

very nice, though I'll never get to see the awesome loading screen on my computer again :/